FreeLoRA: Enabling Training-Free LoRA Fusion for Autoregressive Multi-Subject Personalization

TL;DR

FreeLoRA introduces a training-free method to fuse multiple subject-specific LoRA modules for personalized image generation, simplifying multi-subject customization without re-tuning or complex optimization.

Contribution

It proposes a novel training-free fusion framework for multi-subject personalization using subject-specific LoRA modules and subject-aware inference, avoiding re-tuning.

Findings

Effective multi-subject personalization with high fidelity.

Mitigates overfitting and mutual interference.

Strong performance in subject fidelity and prompt consistency.

Abstract

Subject-driven image generation plays a crucial role in applications such as virtual try-on and poster design. Existing approaches typically fine-tune pretrained generative models or apply LoRA-based adaptations for individual subjects. However, these methods struggle with multi-subject personalization, as combining independently adapted modules often requires complex re-tuning or joint optimization. We present FreeLoRA, a simple and generalizable framework that enables training-free fusion of subject-specific LoRA modules for multi-subject personalization. Each LoRA module is adapted on a few images of a specific subject using a Full Token Tuning strategy, where it is applied across all tokens in the prompt to encourage weakly supervised token-content alignment. At inference, we adopt Subject-Aware Inference, activating each module only on its corresponding subject tokens. This enables training-free fusion of multiple personalized subjects within a single image, while mitigating overfitting and mutual interference between subjects. Extensive experiments show that FreeLoRA achieves strong performance in both subject fidelity and prompt consistency.